Download presentation
Published byRory Hille Modified over 10 years ago
1
Tutorial 1: Sensitivity analysis of an analytical function
2
Example: Analytical nonlinear function
Additive linear and nonlinear terms and one coupling term Contribution to the output variance (reference values): X1: 18.0%, X2: 30.6%, X3: 64.3%, X4: 0.7%, X5: 0.2% Tutorial 1: Sensitivity Analysis
3
Task description Parameterization of the problem Defining DOE scheme
Evaluation of DOE designs Statistical post-processing of DOE Approximation post-processing of DOE Defining MOP search algorithm Evaluation of MOP workflow Statistical post-processing of MOP Approximation post-processing of MOP Reload results in Result Monitoring Use Matlab as solver Use Excel as solver Use Excel plug-in to export data in optiSLang format Tutorial 1: Sensitivity Analysis
4
Project manager Open the project manager Define project name
1. 2. 3. Open the project manager Define project name Create a new project directory Copy optiSLang examples/Coupled_Function into project directory Tutorial 1: Sensitivity Analysis
5
Parameterization of the problem
1. 2. 3. 4. 5. Start a new parametrize workflow (double click) Define workflow name Create a new problem specification Enter problem file name Tutorial 1: Sensitivity Analysis
6
Parameterization of the problem
1. 2. 3. Click “open file” icon in parametrize editor Browse for the SLang input file coupled_function.s Choose file type as INPUT Tutorial 1: Sensitivity Analysis
7
Parameterization of the problem
1. 2. 3. Mark value of X1 in the input file Define X1 as input parameter Enter parameter name Tutorial 1: Sensitivity Analysis
8
Parameterization of the problem
1. 2. Open parameter in parameter three Enter lower and upper bounds Set as default for other variables and repeat for X2 … X5 3. Tutorial 1: Sensitivity Analysis
9
Parameterization of the problem
1. 2. 3. Click “open file” icon in parametrize editor Browse for the SLang output file coupled_solution.s Choose file type as OUTPUT Tutorial 1: Sensitivity Analysis
10
Parameterization of the problem
4. 2. 3. 1. Mark output value in editor Define Y as output parameter Enter parameter name Close parametrize editor Tutorial 1: Sensitivity Analysis
11
Parameterization of the problem
1. 2. 3. Check parameter overview for inputs Check parameter overview for outputs Close overview Tutorial 1: Sensitivity Analysis
12
Define Design Of Experiments (DOE)
2. 1. 3. 4. Start a new DOE workflow (double click) Define workflow name Define workflow identifier (working directory) Enter problem file name Tutorial 1: Sensitivity Analysis
13
Define Design Of Experiments (DOE)
1. 2. 3. 4. Enter solver call (slang –b coupled_function.s) Enter number of parallel runs Choose if design directories should be deleted Start DOE workflow Tutorial 1: Sensitivity Analysis
14
Generate DOE scheme Choose Latin hypercube sampling
1. 2. 3. Choose Latin hypercube sampling Enter number of samples (50…100) Generate samples Close dialog and show samples 4. Tutorial 1: Sensitivity Analysis
15
Generate DOE scheme Start evaluation of samples 1.
Tutorial 1: Sensitivity Analysis
16
Statistics post-processing
3. 1. 5. 2. 4. 6. Linear correlation matrix (In-In, In-Out, Out-In and Out-Out) Quadratic correlation matrix (total values or difference to linear) Histogram of input/output (select variable in 1.) Anthill plot (select variables in 1.) CoD/CoI values (linear: select in 1., quadratic: select in 2.) Ranked linear or quadratic correlations of single response Tutorial 1: Sensitivity Analysis
17
Statistics post-processing
1. 1. 2. Switch between CoD/CoI visualization Extended correlation matrix (optiSLang 3.2) Tutorial 1: Sensitivity Analysis
18
Statistics post-processing
1. 2. Statistical properties of single variable Traffic light plot of response for given safety & failure limit (optiSLang 3.2) Tutorial 1: Sensitivity Analysis
19
Statistics post-processing
1. 2. 3. Show development of correlation coefficients Show design table Export DOE to Excel Tutorial 1: Sensitivity Analysis
20
Statistics post-processing
1. 2. Principal Component Analysis (PCA) of linear correlations Parallel coordinates plot to show designs having an input/output within certain lower and upper bounds Tutorial 1: Sensitivity Analysis
21
Statistics post-processing
1. 2. Significance filter for CoD/CoI Manual filter for CoD/CoI Tutorial 1: Sensitivity Analysis
22
Approximation post-processing
3. 1. 2. 4a. 4b. Anthill plot of parameter 1 and the response Contour plot of approximation function in terms of parameter 1 and 2 (remaining are set to their mean) vs. the response Surface plot of approximation function Details about approximation settings and properties Tutorial 1: Sensitivity Analysis
23
Approximation post-processing
3. 4a. 4b. Manual approximation settings: Parameter subspace Polynomial or MLS (exponential or regularized) Basis polynomial, constant or density dependent influence Transformation settings Tutorial 1: Sensitivity Analysis
24
Meta-Model of Optimal Prognosis (MOP)
2. 1. 3. 4. 5. Start a new MOP workflow (double click) Define workflow name Define workflow identifier (working directory) Choose DOE result file Choose optional problem file Tutorial 1: Sensitivity Analysis
25
Meta-Model of Optimal Prognosis (MOP)
1. 4. 2. 3. 5. CoP settings (sample splitting or cross validation) Investigated approximation models DCoP - accepted reduction in prediction quality to simplify model Filter settings Selection of inputs and outputs Tutorial 1: Sensitivity Analysis
26
Meta-Model of Optimal Prognosis (MOP)
optiSLang console gives detailed information about the investigated models and obtained optimal CoP values Tutorial 1: Sensitivity Analysis
27
Meta-Model of Optimal Prognosis (MOP)
Approximation post-processing automatically shows surface and contour plot of the two most important variables vs. the response CoP values for single variables are shown Tutorial 1: Sensitivity Analysis
28
Overview of different significance values
MOP/CoP close to reference values (detects optimal subspace automatically, represents nonlinear and coupling terms) CoD, k=5 (all inputs) CoI, k=5 (all inputs) CoI, k=3 (manual) CoP, k=3 (automatic) Reference Full model 75% 74% 97% 100% X1 2% 14% 18% X2 30% 28% 31% X3 41% 34% 39% 62% 64% X4 0% - 0.7% X5 1% 0.2% Tutorial 1: Sensitivity Analysis
29
Reload DOE or MOP in Result Monitoring
2. 1. 3. Start a new Results Monitoring workflow (double click) Define workflow name Choose DOE or MOP result file Start Post-Processing Tutorial 1: Sensitivity Analysis
30
Tutorial 1: Use Matlab as solver
31
Use Matlab as solver Matlab input file: coupled_function.m
1. 2. 3. 4. Matlab input file: coupled_function.m Input parameter definition Function evaluation Writing the result file Exit Matlab execution! Tutorial 1: Sensitivity Analysis
32
Use Matlab as solver Call Matlab from Windows: matlab_windows.bat
1. 2. 3. 4. 5. Call Matlab from Windows: matlab_windows.bat Disable splash Disable desktop Disable java virtual machine Minimize remaining command window Wait until Matlab is terminated Tutorial 1: Sensitivity Analysis
33
Use Matlab as solver Call Matlab from Linux: matlab_linux.sh
1. 2. 3. 4. 5. Call Matlab from Linux: matlab_linux.sh Set empty display Disable splash Disable desktop Disable java virtual machine Wait until Matlab is finished Tutorial 1: Sensitivity Analysis
34
Use Matlab as solver 1. 2. Parameterize inputs in optiSLang from coupled_function.m Parameterize output from coupled_solution.txt Tutorial 1: Sensitivity Analysis
35
Use Matlab as solver 1. 2. Open new DOE workflow and select “Run a script file” Choose the batch script and start DOE process Tutorial 1: Sensitivity Analysis
36
Tutorial 1: Use Excel as solver
37
Use Excel as solver 2. 1. 3. Generate Excel file with all inputs in one row and all outputs in one column Mark first input as inputParams Mark first output as outputParams Tutorial 1: Sensitivity Analysis
38
Use Excel as solver Show Macros Enter Macro name Create Macro 1. 2. 3.
Tutorial 1: Sensitivity Analysis
39
Use Excel as solver 1. 2. In Visual Basic environment use import file feature Import predefined macro file inout.bas Tutorial 1: Sensitivity Analysis
40
Use Excel as solver inout module should be shown in the module list
1. 1. 2. inout module should be shown in the module list Delete the empty default module Tutorial 1: Sensitivity Analysis
41
Use Excel as solver The visual basic macro Input file name
1. 2. The visual basic macro Input file name Output file name Tutorial 1: Sensitivity Analysis
42
Use Excel as solver Java script to run Excel in batch mode
1. Java script to run Excel in batch mode Excel file name Tutorial 1: Sensitivity Analysis
43
Use Excel as solver Batch script to run Excel java script
1. Batch script to run Excel java script Call of java script with full path, modify path if necessary! Tutorial 1: Sensitivity Analysis
44
Use Excel as solver Parameterize inputs in optiSLang from input.txt
1. 2. Parameterize inputs in optiSLang from input.txt Parameterize output from output.txt Tutorial 1: Sensitivity Analysis
45
Use Excel as solver 1. 2. Open new DOE workflow and select “Run a script file” Choose the batch script and start DOE process Tutorial 1: Sensitivity Analysis
46
Tutorial 1: Use Excel plug-in
47
Use Excel plug-in Start the plug-in in Excel
1. 2. 3b. 3a. Start the plug-in in Excel Mark input data including parameter names Check parameter names and data array Tutorial 1: Sensitivity Analysis
48
Use Excel plug-in Mark output data including parameter names
2b. 1. 2a. Mark output data including parameter names Check parameter names and data array Tutorial 1: Sensitivity Analysis
49
Use Excel plug-in Choose design numbers
1. 2. Choose design numbers Finish and save data in optiSLang *.bin file Open *.bin in result monitoring workflow Tutorial 1: Sensitivity Analysis
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.